Method and system for the automated delineation of lung regions and costophrenic angles in chest radiographs

Abstract

A method, system, and computer product for the automated segmentation of the lung fields and costophrenic angle (CP) regions in posteroanterior (PA) chest radiographs, wherein image segmentation based on gray-level threshold analysis is performed by applying an iterative global gray-level thresholding method to a chest image based on the features of a global gray-level histogram. Features of the regions in a binary image constructed at each iteration are identified and analyzed to exclude regions external to the lung fields. The initial lung contours that result from this global process are used to facilitate a local gray-level thresholding method. Individual regions-of-interest (ROIs) are placed along the initial contour. A procedure is implemented to determine the gray-level thresholds to be applied to the pixels within the individual ROIs. The result is a binary image, from which final contours are constructed. Smoothing processes are applied, including a unique adaptation of a rolling ball method. CP angles are identified and delineated by using the lung segmentation contours as a means of placing ROIs that capture the CP angle regions. Contrast-based information is employed on a column-by-column basis to identify initial diaphragm points, and maximum gray-level information is used on a row-by-row basis to identify initial costal points. Analysis of initial diaphragm and costal points allows for appropriate adjustment of CP angle ROI positioning. Polynomial curve-fitting is used to combine the diaphragm and costal points into a continuous, smooth CP angle delineation. This delineation is then spliced into the final lung segmentation contours. In addition, quantitative information derived from the CP angle delineations is used to assess the presence of abnormal CP angles.

Dependencies are inferred from the claim text.